Frequency response tests have now become accepted as a sound alternati
ve to sudden tests for the determination of for synchronous machines f
or transient studies. The standard approach to the extraction of machi
ne parameters from the results of frequency response tests generally c
oncentrate on curve-fitting techniques to match the measured magnitude
and phase with a set of time constants. These processes are fraught w
ith difficulty in respect of needing to: first, define the order of th
e model before the analysis can commence; and secondly, initiate the c
urve fitting with initial estimates of the parameters. Unfortunately,
there is not a unique set of time constants which produce a frequency
response measured and a better 'blind' numerical method is therefore n
eeded. The author presents the application of standard linear systems
theory to predict the positions of the poles and zeros in the frequenc
y response and to determine the order of the equivalent circuit requir
ed to model the machine accurately. The process breaks quite naturally
into two parts; the extraction of the time constants from the frequen
cy response, and the determination of the parameters of the equivalent
circuit from those time constants. The basis for the measurement tech
nique is reviewed and the effects of different levels of complexity of
the equivalent circuit are considered with respect to the increased d
ifficulty in extracting the parameters. The ease with which this metho
d copes with the higher-order models and the sequential nature of the
process, working from the lowest frequency to the highest frequency in
the frequency response, justify accepting the procedure. Results obta
ined from tests on production machines are used to illustrate the proc
edures for both time constant extraction and equivalent circuit parame
ter determination to confirm the capabilities of the methods.